Cost and CO2 emissions co-optimisation of green hydrogen production in a grid-connected renewable energy system (2404.11995v1)

Abstract: Green hydrogen is essential for producing renewable fuels that are needed in sectors that are hard to electrify directly. Hydrogen production in a grid-connected hybrid renewable energy plant necessitates smart planning to meet long-term hydrogen trading agreements while minimising costs and emissions. Previous research analysed economic and environmental impact of hydrogen production based on full foresight of renewable energy availabilty, electricity price, and CO2 intensity in the electricity grid. However, the full foresight assumption is impractical in day-to-day operation, often leading to underestimations of both the cost and CO2 emissions associated with hydrogen production. Therefore, this research introduces a novel long-term planner that uses historical data and short-term forecasts to plan hydrogen production in the day-to-day operation of a grid-connected hybrid renewable energy plant. The long-term planner co-minimises cost and CO2 emissions to determine the hydrogen production for the next day taking into account the remaining hydrogen production and the time remaining until the end of the delivery period, which can be a week, a month, or a year. Extended delivery periods provide operation flexibility, enabling cost and CO2 emissions reductions. Significant reductions in CO2 emissions can be achieved with relatively small increases in the levelised cost. Under day-to-day operation, the levelised cost of hydrogen is marginally higher than that of the full foresight; the CO2 emissions can be up to 60% higher. Despite a significant portion of the produced hydrogen not meeting the criteria for green hydrogen designation under current rules, CO2 emissions are lower than those from existing alternative hydrogen production methods. These results underscore the importance of balancing cost considerations with environmental impacts in operational decision-making.

• The paper proposes a novel dual-planner strategy that combines long-term historical data with short-term forecasts to plan hydrogen production.
• The optimization model minimizes a weighted sum of electricity, operational, and CO2 costs under realistic grid-connected conditions.
• Results reveal a significant trade-off where strategies lowering CO2 may increase cost, exposing limitations in full foresight assumptions.

This paper (Cost and CO2 emissions co-optimisation of green hydrogen production in a grid-connected renewable energy system, 18 Apr 2024) addresses the practical challenge of planning green hydrogen production in grid-connected hybrid renewable energy systems under realistic day-to-day operation conditions, contrasting it with often-used, but impractical, full foresight assumptions. The core problem is how to meet long-term hydrogen delivery commitments while simultaneously minimizing both production costs and CO2 emissions from electricity imports, given that detailed forecasts for renewable energy availability, electricity prices, and grid CO2 intensity are only reliable for short horizons.

The research proposes a novel approach for long-term planning in a day-to-day operational setting. Instead of relying on unreliable long-term forecasts, the proposed long-term planner uses a combination of historical data and available short-term forecasts. This planner works in conjunction with a daily planner to determine the optimal hydrogen production schedule.

System Architecture and Control

The studied system consists of a grid-connected hydrogen production plant including solar PV, wind turbines, an electrolyser, and an inverter. The control logic involves two main planners:

1. Long-Term Planner: Executed daily, it determines the total mass of hydrogen ($\hat{M}_{d+1}$) to be produced on the upcoming day ($d+1$). Its novelty lies in using a time window that combines recent historical data (actual values) and short-term forecasts. This window shrinks daily, adapting to the remaining hydrogen target and time until the end of the overall delivery period (which can be a day, week, month, or year). The objective is to co-minimize cost and CO2 emissions over this shrinking historical+forecast window, ensuring the long-term target is met without end-of-period corrections.
2. Daily Planner: Also executed daily, it takes the upcoming day's target mass ($\hat{M}_{d+1}$) from the long-term planner and creates an hour-by-hour optimal production plan ($\hat{m}_{d+1, t}$) for the upcoming day, extending into the first few hours of the following day to avoid suboptimal end-of-day decisions. This optimization uses short-term forecasts for solar/wind availability, electricity prices, and CO2 intensity.
3. Hydrogen Production Plant: Executes the hourly plan ($\hat{m}_{d+1, t}$) determined by the daily planner. It manages the power flows (from renewables, imported/exported from the grid) in real-time to meet the planned hydrogen production, prioritizing production over grid power commitments if forecasts don't match reality.

Optimization Model

The core of the planners and the plant model is a co-optimization objective function that minimizes a weighted sum of CO2 emissions cost and electricity/operation costs:

$$\min \hat {C}_{\alpha} = \alpha\hat{C}_{CO_2} + (1-\alpha)(\hat{C}_{e} + \hat{C}_o)$$

where $\alpha$ is a weighting factor ($0 \le \alpha \le 1$) allowing for balancing cost and emissions priorities. The model includes standard power flow balances (DC and AC buses), renewable curtailment options, grid connection limits, electrolyser energy balance, power limits, and ramping constraints. The electrolyser efficiency is modeled based on its capacity factor. To maintain linearity for efficient solving with commercial solvers like Gurobi, the hydrogen power output is treated as a decision variable constrained by the power input and efficiency, with the capacity factor-efficiency product being approximated as a linear function based on empirical data.

Practical Results and Implications

The paper compares the day-to-day operation (using the proposed planners) with a benchmark using full annual foresight, for different delivery periods (daily, weekly, monthly, yearly) using data from Denmark (2018-2021).

• Cost-Emissions Trade-off: Both benchmark and day-to-day operations show a trade-off: reducing specific CO2 emissions generally increases the levelised cost of hydrogen (LCOH). Longer delivery periods provide more flexibility for optimizing this trade-off.
• Day-to-Day vs. Benchmark:
  • LCOH is only marginally higher (slightly underestimated by full foresight) under day-to-day operation compared to the benchmark, especially when prioritizing cost minimization.
  • Specific CO2 emissions can be significantly higher (up to 60% higher) under day-to-day operation compared to the benchmark, especially when prioritizing emissions minimization. This is a crucial practical finding: models assuming full foresight significantly underestimate the environmental impact of green hydrogen production in realistic operating conditions.
  • For yearly delivery targets, prioritizing cost minimization in day-to-day operation could paradoxically lead to higher overall costs towards the end of the year compared to co-minimization, due to the operational constraint of meeting the long-term target forcing continuous production when less favorable conditions (high price/emissions) arise.
• Green Hydrogen Evaluation: Evaluating the produced hydrogen against current EU regulations [Directorate_General_for_Energy]:
  • Even with high renewable penetration in the grid, a significant portion (approx. 30% in the Danish case paper) of the produced hydrogen may not qualify as "green" under current rules based on annual averages or price thresholds.
  • However, even this "non-green" hydrogen produced through the optimized system has significantly lower CO2 emissions per kg compared to conventional methods like Steam Methane Reforming (SMR) or coal gasification.
  • The current regulations are criticized for a lack of transparency and effectiveness in linking green status directly to CO2 emissions, as price correlation with emissions is weak, and annual averages don't capture hourly variability.

Recommendations for Green Hydrogen Regulations

Based on the practical findings, the paper suggests improvements to green hydrogen regulations:

1. Hourly CO2 Accounting: Adopt a transparent system based on the actual hourly CO2 intensity of the electricity used for production, rather than annual averages or price signals.
2. Lower CO2 Threshold: Reduce the specific CO2 emission threshold for hydrogen to be labeled "green" (currently equivalent to approx. 3.6 kg CO2/kg H2 under one rule), suggesting a lower limit like 1.5 kg CO2/kg H2 as being achievable and more environmentally sound.

Implementation Considerations

Implementing this approach in practice requires:

• Access to reliable short-term forecasts (e.g., for solar/wind, price, CO2 intensity). While the paper assumes perfect forecasts for simplicity, real-world implementation needs robust forecasting models.
• A robust optimization solver capable of handling linear programming problems with a significant number of variables and constraints over the chosen forecast horizon (e.g., Gurobi, as used in the paper).
• Infrastructure for collecting and processing historical data for the long-term planner.
• Real-time data streams from the plant and grid for operational decisions and updating the historical data used by the long-term planner.
• The computational load for running the optimization daily needs to be considered, although the described model appears to be a linear program, which is computationally tractable.
• The ramp rate constraints of the electrolyser are critical operational limits that must be accurately modeled.

The research highlights that realistic planning for green hydrogen must account for the limitations of foresight and the significant impact this has on achieving emissions targets, suggesting that current planning methods based on full foresight may lead to overoptimistic environmental assessments. The proposed planning approach offers a practical step towards more realistic operational planning for grid-connected hydrogen plants.